Towards experimental evaluation of code obfuscation techniques

While many obfuscation schemes proposed, none of them satisfy any strong definition of obfuscation. Furthermore secure general-purpose obfuscation algorithms have been proven to be impossible. Nevertheless, obfuscation schemes which in practice slow down malicious reverse-engineering by obstructing code comprehension for even short periods of time are considered a useful protection against malicious reverse engineering. In previous works, the difficulty of reverse engineering has been mainly estimated by means of code metrics, by the computational complexity of static analysis or by comparing the output of de-obfuscating tools. In this paper we take a different approach and assess the difficulty attackers have in understanding and modifying obfuscated code through controlled experiments involving human subjects